Few-shot Image Classification on tieredImagenet (test)
98.1AccuracyCAML
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| CAMLWays=5, Shots=5, Setting=Universal Meta-Learning2023.10 | 98.1 | — | — | — | |
| ProtoNetWays=5, Shots=5, Setting=Universal Meta-Learning, Backbone Status=Unfrozen2023.10 | 97.4 | — | — | — | |
| P>M>FWays=5, Shots=5, Setting=In-Domain [Meta-Training], Backbone=ViT-base2023.10 | 97.3 | — | — | — | |
| SNAILWays=5, Shots=5, Setting=Universal Meta-Learning2023.10 | 97.3 | — | — | — | |
| GPICLWays=5, Shots=5, Setting=Universal Meta-Learning2023.10 | 97.2 | — | — | — | |
| MetaQDAWays=5, Shots=5, Setting=Universal Meta-Learning2023.10 | 97 | — | — | — | |
| MetaOptWays=5, Shots=5, Setting=Universal Meta-Learning, Backbone Status=Frozen2023.10 | 96.5 | — | — | — | |
| ProtoNetWays=5, Shots=5, Setting=Universal Meta-Learning, Backbone Status=Frozen2023.10 | 96.1 | — | — | — | |
| CAMLWays=5, Shots=1, Setting=Universal Meta-Learning2023.10 | 95.4 | — | — | — | |
| GPICLWays=5, Shots=1, Setting=Universal Meta-Learning2023.10 | 94.6 | — | — | — | |
| P>M>FWays=5, Shots=1, Setting=In-Domain [Meta-Training], Backbone=ViT-base2023.10 | 93.5 | — | — | — | |
| ProtoNetWays=5, Shots=1, Setting=Universal Meta-Learning, Backbone Status=Unfrozen2023.10 | 93.5 | — | — | — | |
| SNAILWays=5, Shots=1, Setting=Universal Meta-Learning2023.10 | 93.1 | — | — | — | |
| MetaOptWays=5, Shots=5, Setting=Universal Meta-Learning, Backbone Status=Unfrozen2023.10 | 89.6 | — | — | — | |
| MetaQDAWays=5, Shots=1, Setting=Universal Meta-Learning2023.10 | 89.4 | — | — | — | |
| ProtoDiffshot=5-shot2023.06 | 88.75 | — | — | — | |
| MetaOptWays=5, Shots=1, Setting=Universal Meta-Learning, Backbone Status=Frozen2023.10 | 88.2 | — | — | — | |
| SetFeatshot=5-shot2023.06 | 87.59 | — | — | — | |
| Meta DeepBDCshot=5-shot2023.06 | 87.31 | — | — | — | |
| ProtoNetWays=5, Shots=1, Setting=Universal Meta-Learning, Backbone Status=Frozen2023.10 | 87.3 | — | — | — | |
| SUNshot=5-shot2023.06 | 86.74 | — | — | — | |
| CANshot=5-shot2023.06 | 84.23 | — | — | — | |
| Meta-Baselineshot=5-shot2023.06 | 83.74 | — | — | — | |
| MetaOptNetshot=5-shot2023.06 | 81.75 | — | — | — | |
| CTMshot=5-shot2023.06 | 81.05 | — | — | — | |
| noHub-SBackbone=ResNet-18, Number of shots=1-shot2023.03 | 80.6 | — | — | 7.67 | |
| noHub-SBackbone=WideResNet28-10, Number of shots=1-shot2023.03 | 80.46 | — | — | 7.67 | |
| TapNetshot=5-shot2023.06 | 80.26 | — | — | — | |
| noHubBackbone=ResNet-18, Number of shots=1-shot2023.03 | 79.77 | — | — | 6.83 | |
| noHubBackbone=WideResNet28-10, Number of shots=1-shot2023.03 | 79.76 | — | — | 7 | |
| TCPRBackbone=ResNet-18, Number of shots=1-shot2023.03 | 77.18 | — | — | 3.33 | |
| EASEBackbone=ResNet-18, Number of shots=1-shot2023.03 | 77.05 | — | — | 4 | |
| CL2Backbone=ResNet-18, Number of shots=1-shot2023.03 | 76.97 | — | — | 3 | |
| MetaOptWays=5, Shots=1, Setting=Universal Meta-Learning, Backbone Status=Unfrozen2023.10 | 76.6 | — | — | — | |
| EASEBackbone=WideResNet28-10, Number of shots=1-shot2023.03 | 76.59 | — | — | 3.67 | |
| TCPRBackbone=WideResNet28-10, Number of shots=1-shot2023.03 | 76.51 | — | — | 4 | |
| L2Backbone=WideResNet28-10, Number of shots=1-shot2023.03 | 76.19 | — | — | 2.67 | |
| ProtoDiffshot=1-shot2023.06 | 75.97 | — | — | — | |
| L2Backbone=ResNet-18, Number of shots=1-shot2023.03 | 75.94 | — | — | 2.17 | |
| CL2Backbone=WideResNet28-10, Number of shots=1-shot2023.03 | 75.17 | — | — | 2 | |
| GAPBackbone=Conv-4, N-way=5-way, K-shot=5-shot2023.04 | 74.9 | — | — | — | |
| Approximate GAPBackbone=Conv-4, N-way=5-way, K-shot=5-shot2023.04 | 74.41 | — | — | — | |
| PAMELABackbone=Conv-4, N-way=5-way, K-shot=5-shot2023.04 | 74.39 | — | — | — | |
| WarpGradBackbone=Conv-4, N-way=5-way, K-shot=5-shot2023.04 | 74.1 | — | — | — | |
| SetFeatshot=1-shot2023.06 | 73.63 | — | — | — | |
| CxGradBackbone=Conv-4, N-way=5-way, K-shot=5-shot2023.04 | 73.55 | — | — | — | |
| L2FBackbone=Conv-4, N-way=5-way, K-shot=5-shot2023.04 | 73.34 | — | — | — | |
| SUNshot=1-shot2023.06 | 72.99 | — | — | — | |
| Meta DeepBDCshot=1-shot2023.06 | 72.34 | — | — | — | |
| ReRepBackbone=WideResNet28-10, Number of shots=1-shot2023.03 | 71.83 | — | — | 3.17 | |
| NoneBackbone=WideResNet28-10, Number of shots=1-shot2023.03 | 71.29 | — | — | 0.83 | |
| ALFABackbone=Conv-4, N-way=5-way, K-shot=5-shot2023.04 | 70.54 | — | — | — | |
| METALBackbone=Conv-4, N-way=5-way, K-shot=5-shot2023.04 | 70.4 | — | — | — | |
| MAMLBackbone=Conv-4, N-way=5-way, K-shot=5-shot2023.04 | 70.3 | — | — | — | |
| Sparse MAML++Backbone=Conv-4, N-way=5-way, K-shot=5-shot2023.04 | 69.92 | — | — | — | |
| CANshot=1-shot2023.06 | 69.89 | — | — | — | |
| BOILBackbone=Conv-4, N-way=5-way, K-shot=5-shot2023.04 | 69.37 | — | — | — | |
| Meta-SGD+Backbone=Conv-4, N-way=5-way, K-shot=5-shot2023.04 | 69.28 | — | — | — | |
| Sparse-ReLU-MAMLBackbone=Conv-4, N-way=5-way, K-shot=5-shot2023.04 | 69.06 | — | — | — | |
| Sparse MAMLBackbone=Conv-4, N-way=5-way, K-shot=5-shot2023.04 | 68.83 | — | — | — | |
| Meta-Baselineshot=1-shot2023.06 | 68.62 | — | — | — | |
| ReRepBackbone=ResNet-18, Number of shots=1-shot2023.03 | 67.07 | — | — | 3.67 | |
| ZNBackbone=ResNet-18, Number of shots=1-shot2023.03 | 66.21 | — | — | 2.5 | |
| MetaOptNetshot=1-shot2023.06 | 65.81 | — | — | — | |
| ZNBackbone=WideResNet28-10, Number of shots=1-shot2023.03 | 65.64 | — | — | 2.5 | |
| CTMshot=1-shot2023.06 | 64.78 | — | — | — | |
| ECMLBackbone=Conv-4, N-way=5-way, K-shot=5-shot2023.04 | 64.77 | — | — | — | |
| TapNetshot=1-shot2023.06 | 63.08 | — | — | — | |
| NoneBackbone=ResNet-18, Number of shots=1-shot2023.03 | 62.61 | — | — | 0 | |
| GAPBackbone=Conv-4, N-way=5-way, K-shot=1-shot2023.04 | 57.6 | — | — | — | |
| WarpGradBackbone=Conv-4, N-way=5-way, K-shot=1-shot2023.04 | 57.2 | — | — | — | |
| Approximate GAPBackbone=Conv-4, N-way=5-way, K-shot=1-shot2023.04 | 56.86 | — | — | — | |
| CxGradBackbone=Conv-4, N-way=5-way, K-shot=1-shot2023.04 | 55.55 | — | — | — | |
| PAMELABackbone=Conv-4, N-way=5-way, K-shot=1-shot2023.04 | 54.81 | — | — | — | |
| L2FBackbone=Conv-4, N-way=5-way, K-shot=1-shot2023.04 | 54.4 | — | — | — | |
| METALBackbone=Conv-4, N-way=5-way, K-shot=1-shot2023.04 | 54.34 | — | — | — | |
| Sparse MAML++Backbone=Conv-4, N-way=5-way, K-shot=1-shot2023.04 | 53.91 | — | — | — | |
| Sparse MAMLBackbone=Conv-4, N-way=5-way, K-shot=1-shot2023.04 | 53.47 | — | — | — | |
| Sparse-ReLU-MAMLBackbone=Conv-4, N-way=5-way, K-shot=1-shot2023.04 | 53.18 | — | — | — | |
| ALFABackbone=Conv-4, N-way=5-way, K-shot=1-shot2023.04 | 53.16 | — | — | — | |
| ARMLBackbone=Conv-4, N-way=5-way, K-shot=1-shot2023.04 | 52.91 | — | — | — | |
| MT-netBackbone=Conv-4, N-way=5-way, K-shot=1-shot2023.04 | 51.95 | — | — | — | |
| MAMLBackbone=Conv-4, N-way=5-way, K-shot=1-shot2023.04 | 51.7 | — | — | — | |
| Meta-SGD+Backbone=Conv-4, N-way=5-way, K-shot=1-shot2023.04 | 50.92 | — | — | — | |
| BOILBackbone=Conv-4, N-way=5-way, K-shot=1-shot2023.04 | 48.58 | — | — | — | |
| ECMLBackbone=Conv-4, N-way=5-way, K-shot=1-shot2023.04 | 47.34 | — | — | — | |
| Assoc-AlignBackbone=WRN, Approach=transductive feature extraction2022.08 | — | 74.4 | 86.61 | — | |
| BOILBackbone=Conv4, Approach=inductive inference2022.08 | — | 66.45 | 69.37 | — | |
| DPGNBackbone=RN12, Approach=transductive feature extraction and inference2022.08 | — | 72.45 | 87.24 | — | |
| FRNBackbone=RN12, Approach=inductive inference2022.08 | — | 71.16 | 86.01 | — | |
| LaplacianShotBackbone=WRN, Approach=transductive inference2022.08 | — | 80.22 | 87.49 | — | |
| LIF-EMDBackbone=RN12, Approach=transductive feature extraction and inference2022.08 | — | 73.76 | 87.83 | — | |
| MAMLBackbone=Conv4, Approach=inductive inference2022.08 | — | 51.67 | 70.3 | — | |
| MetaQDABackbone=WRN, Approach=transductive feature extraction2022.08 | — | 74.33 | 89.56 | — | |
| PALBackbone=RN12, Approach=transductive feature extraction and inference2022.08 | — | 72.25 | 86.95 | — | |
| PEME-BMSBackbone=WRN, Approach=transductive feature extraction and inference2022.08 | — | 86.07 | 91.09 | — | |
| Proto-CompletionBackbone=RN12, Approach=transductive feature extraction and inference2022.08 | — | 81.04 | 87.42 | — | |
| PT+MAPBackbone=WRN, Approach=transductive feature extraction and inference2022.08 | — | 85.67 | 90.45 | — | |
| ReRankBackbone=WRN, Approach=transductive feature extraction and inference2022.08 | — | 79.5 | 84.8 | — | |
| S2M2Backbone=WRN, Approach=transductive feature extraction2022.08 | — | 73.71 | 88.59 | — |